From a clinical question to an analysis-ready cohort Clinical data analysis begins before the first statistical test. A clear question, a well-defined unit of observation, and a set of logical checks are more important than a long list of models. This series follows patients who underwent VATS wedge resection with pleural abrasion for primary spontaneous... Continue Reading →

Sensitivity, co-expression, and planning the next study The final part asks how much the synthetic conclusions depend on simple analysis choices. It also compares the relationship between the two miRNAs across groups and shows a prospective power calculation. The data and results remain fully synthetic. Sensitivity analysis A sensitivity analysis repeats the primary comparison under... Continue Reading →

Adjusted models and cross-validated discrimination This part asks two related questions. First, are the synthetic miRNA measurements associated with case status after adjustment for basic covariates? Second, how well do the measurements separate generated acne and control records in cross-validation? These are association and discrimination questions. They do not establish causation, diagnosis, or clinical utility.... Continue Reading →

Clinical associations and exploratory subgroups The primary analysis compared synthetic acne and control records. This part stays within the generated acne group and asks if miRNA ΔCq values vary with clinical features. All data are synthetic and no result is medical evidence. Secondary analyses are easy to overread. Many tests create more chances for a... Continue Reading →

From Cq values to the primary miRNA result This post uses the synthetic cohort introduced in Part 1. Every measurement and result is generated for teaching. The analysis has no clinical or diagnostic meaning. The main challenge in this section is interpretation. qPCR produces Cq values. The analysis uses a reference measurement to turn them... Continue Reading →

Building a synthetic dermatology miRNA study This is the first post in a five-part series about a small clinical data science workflow. The example studies two miRNA measurements in acne and control groups. Every record is synthetic. The numbers were generated from random distributions with a fixed seed. They do not come from patients, a... Continue Reading →

process is satisfying.

Recognizing exhaustion is not enough to stop

to pause everything and focus on what matters most.

is a necessity for some of our thoughts about 'conventional' assumptions.